EA Portfolio Lab
Portfolio Research
A tool for thinking about how systems combine — exposure, correlation and blended behaviour, computed from the same backtest records as everything else on this site. It is a research instrument, not investment advice, and nothing here is a recommended portfolio.
This page does not recommend any portfolio, allocation or system. It computes arithmetic over historical backtest records under stated assumptions. Whether any combination suits your circumstances is a judgement this tool cannot make.
01Catalogue exposure
Where the 14 published systems sit — the concentration a combination would inherit before any selection is made.
By market
- Forex 7 · 50%
- Crypto 4 · 28.6%
- Gold 3 · 21.4%
By strategy
- Scalping 13 · 92.9%
- Martingale 1 · 7.1%
By risk level
- High 5 · 35.7%
- Medium 5 · 35.7%
- Low 4 · 28.6%
02Correlation matrix
Pearson correlation of monthly returns over the months each pair actually shares. Pairs with fewer than 24 common months show “—”, because a correlation from a handful of shared months is noise wearing a number.
| JOKER | JAYRO | GRANDSLAM | GODSPEED | GALOA | COSMOS | KOKOMO | METRO | MIRUMO | SENA | RINA | NEXUS | EVERGREEN | SUPREMACY | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| JOKER | .23 | .03 | .14 | -.02 | .09 | .21 | .09 | .02 | .08 | -.01 | .12 | -.08 | .03 | |
| JAYRO | .23 | -.08 | .38 | — | -.07 | — | — | — | — | — | -.05 | — | -.01 | |
| GRANDSLAM | .03 | -.08 | -.56 | — | .03 | — | — | .04 | — | — | .37 | — | .22 | |
| GODSPEED | .14 | .38 | -.56 | .09 | -.03 | .10 | .19 | .06 | .08 | .16 | -.10 | .10 | .03 | |
| GALOA | -.02 | — | — | .09 | -.06 | .19 | .06 | .24 | .09 | .13 | .23 | -.08 | .27 | |
| COSMOS | .09 | -.07 | .03 | -.03 | -.06 | -.06 | .04 | -.20 | .19 | .07 | .11 | -.12 | -.14 | |
| KOKOMO | .21 | — | — | .10 | .19 | -.06 | .01 | .27 | .48 | .28 | -.12 | .07 | .02 | |
| METRO | .09 | — | — | .19 | .06 | .04 | .01 | -.05 | .23 | .20 | .01 | .11 | -.01 | |
| MIRUMO | .02 | — | .04 | .06 | .24 | -.20 | .27 | -.05 | .16 | .14 | .10 | -.06 | -.02 | |
| SENA | .08 | — | — | .08 | .09 | .19 | .48 | .23 | .16 | .49 | .09 | .27 | .10 | |
| RINA | -.01 | — | — | .16 | .13 | .07 | .28 | .20 | .14 | .49 | -.09 | .16 | -.19 | |
| NEXUS | .12 | -.05 | .37 | -.10 | .23 | .11 | -.12 | .01 | .10 | .09 | -.09 | .03 | -.05 | |
| EVERGREEN | -.08 | — | — | .10 | -.08 | -.12 | .07 | .11 | -.06 | .27 | .16 | .03 | -.10 | |
| SUPREMACY | .03 | -.01 | .22 | .03 | .27 | -.14 | .02 | -.01 | -.02 | .10 | -.19 | -.05 | -.10 |
These are correlations of backtest months, and they describe how two test records moved historically. Correlation in a backtest window does not promise the same relationship in the future.
03Combination builder
Pick two to four systems, choose a weighting scheme, and see how the blend of their monthly returns behaved over the window they all share. Everything is computed on backtest monthly returns only.
Select at least two systems.
The selected systems share fewer than 12 months of common history — not enough to say anything.
- Common window—
- Shared months—
- Average pairwise correlation—
- Positive months—
- Mean month—
- Worst month—
- Volatility (annualised)—
- Max drawdown (monthly closes)—
- Return / max DD—
- CAGR—
- Cumulative return—
Risk-parity weighting did not converge for this selection; inverse-volatility weights are shown instead.
Each system over the same window
| Systems | Weight | Positive months | Worst month | Max drawdown (monthly closes) | Cumulative return |
|---|
Computed from backtest monthly returns over the common window only, with the selected weighting held fixed and rebalanced monthly. Drawdown here is measured on monthly closing values, which understates intra-month declines. Historical arithmetic, not a forecast.
Inverse-volatility and risk-parity weights are computed from the same window they are then applied to — an in-sample construction. It shows how a weighting scheme shapes the blend, not a result anyone could have obtained in advance.
Rolling correlation
How each selected pair moved together over time — computed over the same common window as the blend above.
Not enough shared months for this window.
Shorter windows are noisier: a 12-month correlation swings on a handful of months. The 24-month view is the reference; 12 months is shown for texture, not for conclusions.
Correlation stability (rolling 24 months)
| Pair | Full-window r | Mean | Lowest | Highest |
|---|
A pair whose rolling correlation ranges widely is not reliably diversifying: the full-window number is an average over regimes, not a property of the pair.
Loss & drawdown overlap
| Pair | Joint loss months | If independent | Both in drawdown |
|---|
Joint loss months counts months where both systems closed negative, next to the share expected if losses were independent (the product of each system’s own loss-month rates). “Both in drawdown” is the share of months both sat below a prior peak of their monthly-close curves within the common window. Backtest arithmetic, not a forecast of joint behaviour.
Systems we have shipped.
Each Expert Advisor is documented with its concept, logic, risk model and test conditions. No black boxes, no unexplained results.